v1.1.2

Bug Fixes

Server (MCP tools)

• Critical: Fixed 11 tools that crashed on non-MCPVideoError exceptions (AI features, transitions, quality checks, edit timeline) — now properly handled instead of calling .to_dict() on plain exceptions
• Added input validation: opacity (0-1), fps (>0), scale_factor (2-16), frame_count (1-100), fade_in/fade_out (>=0)
• Added batch operation whitelist validation (unknown operations now rejected early)
• Exposed rounded_corners parameter in video_layout_pip (was hardcoded to False)
• Fixed keyword arg passing in video_layout_pip

Python Client

• Fixed keyword arg mismatches in overlay_video, split_screen, all effect_* methods, add_generated_audio, batch, text_subtitles
• Client methods now correctly pass parameters by name to match engine signatures

CLI

• Fixed audio-waveform dispatch (now passes output_path)
• Fixed export-frames format conversion (jpg → mjpeg for FFmpeg compatibility)
• Fixed auto-chapters to handle both tuple and dict return types
• Added JSON error handling for write-metadata (invalid JSON now prints helpful error)
• Fixed B904 ruff lint error

Engine

• export_frames now accepts "mjpeg" as alias for "jpg"

AI Engine

• Graceful handling when cv2.dnn_superres is unavailable (raises MCPVideoError instead of AttributeError)

Tests

• Fixed test_batch_unknown_operation assertion for new validation response shape
• Fixed test_ai_upscale_missing_dependency for environments without opencv-contrib
• Updated version assertions to 1.1.2